q27 clustering and career goals

## # A tibble: 19 x 2
##      Q29     n
##    <dbl> <int>
##  1    13   915
##  2    NA   771
##  3     5   745
##  4     4   636
##  5     8   351
##  6     3   180
##  7    11   173
##  8    16   151
##  9    10   128
## 10     7   119
## 11    17   114
## 12     1    84
## 13    12    71
## 14    18    69
## 15     6    55
## 16     2    21
## 17    14    13
## 18     9     6
## 19    15     3

## # A tibble: 5 x 2
##     Q32     n
##   <dbl> <int>
## 1     1  1038
## 2     2  1122
## 3     3  1121
## 4     4   702
## 5    NA   622

## # A tibble: 5 x 2
##     Q32     n
##   <dbl> <int>
## 1     1    72
## 2     2    81
## 3     3    80
## 4     4    75
## 5    NA   463

## # A tibble: 13 x 2
##      Q33     n
##    <dbl> <int>
##  1     1  1004
##  2     2    63
##  3     3   143
##  4     4   904
##  5     5    60
##  6     6    68
##  7     7    66
##  8     8   252
##  9     9   419
## 10    10   420
## 11    11   224
## 12    12   388
## 13    NA   594

## # A tibble: 13 x 2
##      Q33     n
##    <dbl> <int>
##  1     1   947
##  2     4   835
##  3    10   399
##  4     9   382
##  5    12   351
##  6     8   237
##  7    11   199
##  8    NA   133
##  9     3   119
## 10     6    62
## 11     5    57
## 12     7    57
## 13     2    56

## # A tibble: 6 x 2
##     Q37     n
##   <dbl> <int>
## 1     1  2985
## 2     2  1016
## 3     3    29
## 4     4    44
## 5  1234     1
## 6    NA   530

Drop NAs for specific questions and filter out disciplines with fewer than 30 (the cutoff) students in sample

## # A tibble: 18 x 2
##      Q29     n
##    <dbl> <int>
##  1    13   644
##  2     5   540
##  3     4   443
##  4     8   243
##  5    16   120
##  6     3   111
##  7    11   111
##  8     7    95
##  9    10    95
## 10    17    81
## 11     1    70
## 12    12    57
## 13    18    42
## 14     6    35
## 15     2    16
## 16    14     8
## 17     9     3
## 18    15     1

## # A tibble: 14 x 2
##      Q29     n
##    <dbl> <int>
##  1    13   644
##  2     5   540
##  3     4   443
##  4     8   243
##  5    16   120
##  6     3   111
##  7    11   111
##  8     7    95
##  9    10    95
## 10    17    81
## 11     1    70
## 12    12    57
## 13    18    42
## 14     6    35

Major counts and percentages

Q29	major	n	pct_total	cumulat_pct
13	Mec	644	23.97	23.97
5	Che	540	20.10	44.06
4	Civ	443	16.49	60.55
8	Ele	243	9.04	69.59
16	Softw	120	4.47	74.06
3	Bio	111	4.13	78.19
11	Ind	111	4.13	82.32
7	Comp	95	3.54	85.86
10	Env/Eco	95	3.54	89.39
17	Str/Arc	81	3.01	92.41
1	Aer/Oce	70	2.61	95.01
12	Mat	57	2.12	97.13
18	Gen	42	1.56	98.70
6	Con	35	1.30	100.00

Gender counts overall

Q37	n	pct_total
1	1973	73.43
2	678	25.23
3	15	0.56
4	21	0.78

fill in 0s for NAs for specific items (Q1, Q3, Q5)

Drop majors with low counts (below 30 students in sample)

## # A tibble: 14 x 2
##      Q29     n
##    <dbl> <int>
##  1    13   618
##  2     5   523
##  3     4   433
##  4     8   240
##  5    16   118
##  6     3   110
##  7    11   108
##  8     7    93
##  9    10    90
## 10    17    78
## 11     1    70
## 12    12    55
## 13    18    40
## 14     6    32

## # A tibble: 14 x 2
##    major       n
##    <chr>   <int>
##  1 Mec       618
##  2 Che       523
##  3 Civ       433
##  4 Ele       240
##  5 Softw     118
##  6 Bio       110
##  7 Ind       108
##  8 Comp       93
##  9 Env/Eco    90
## 10 Str/Arc    78
## 11 Aer/Oce    70
## 12 Mat        55
## 13 Gen        40
## 14 Con        32

Drop majors with NA as major

## # A tibble: 14 x 2
##    major       n
##    <chr>   <int>
##  1 Mec       618
##  2 Che       523
##  3 Civ       433
##  4 Ele       240
##  5 Softw     118
##  6 Bio       110
##  7 Ind       108
##  8 Comp       93
##  9 Env/Eco    90
## 10 Str/Arc    78
## 11 Aer/Oce    70
## 12 Mat        55
## 13 Gen        40
## 14 Con        32

Clustering Process (two-step process using UMAP + HDBSCAN)

First perform dimension reduction using UMAP

## NULL

##           [,1]      [,2]
## [1,]  15.62332 -15.97910
## [2,]  11.12283 -18.44035
## [3,] -34.20802  18.18004

## [1]  15.623323  11.122833 -34.208024 -34.419584   6.551877 -34.582958

## [1] -15.97910 -18.44035  18.18004  18.60377 -23.96864  18.41222

Next, perform clustering with HDBSCAN

## HDBSCAN clustering for 2608 objects.
## Parameters: minPts = 120
## The clustering contains 6 cluster(s) and 241 noise points.
## 
##    0    1    2    3    4    5    6 
##  241 1080  575  175  147  264  126 
## 
## Available fields: cluster, minPts, cluster_scores, membership_prob,
##                   outlier_scores, hc

Join the dataframes back together again

## # A tibble: 7 x 2
##   cluster     n
##     <dbl> <int>
## 1       1  1080
## 2       2   575
## 3       5   264
## 4       0   241
## 5       3   175
## 6       4   147
## 7       6   126

Views of when climate change will affect different groups broken down by cluster assignments

Looking for patterns in the clusters

Understanding cluster compositions more

First create a dataframe with the rankings for each cluster, where a lower ranking means students in the cluster think climate change will affect more categories sooner.

## # A tibble: 7 x 3
##   cluster_time_rank cluster cluster_avg
##               <int>   <dbl>       <dbl>
## 1                 1       1       0.386
## 2                 2       4       3.46 
## 3                 3       0       4.76 
## 4                 4       5       5.14 
## 5                 5       6       7.13 
## 6                 6       2       9.24 
## 7                 7       3      17.4

Set clustering colors for all plots

Look at distribution of cluster for how when they think each community may be affected by global warming

Same information but faceted with clusters

** This is a good plot for seeing that a cluster’s beliefs about effects of global warming on different populations at different times vary in a clear pattern

Overall count for clusters

Broken plot

Career goals analysis

Chi square tests of whether different clusters (students with different temporal discounting patterns of when climate change will affect various groups) want to address climate-related outcomes in their careers –>

*MUTED

Question 2 items (Working in various sectors) and climate change impacts cluster

## # A tibble: 14,462 x 5
##    student_id major cluster_time_rank Q2_item Q2_resp
##         <int> <chr>             <int> <chr>     <dbl>
##  1          1 Ele                   5 Q2a           4
##  2          1 Ele                   5 Q2b           0
##  3          1 Ele                   5 Q2c           4
##  4          1 Ele                   5 Q2d           2
##  5          1 Ele                   5 Q2e           2
##  6          1 Ele                   5 Q2f           2
##  7          1 Ele                   5 Q2g           2
##  8          2 Ele                   4 Q2a           4
##  9          2 Ele                   4 Q2b           1
## 10          2 Ele                   4 Q2c           2
## # ... with 14,452 more rows

Using kruskal-wallis instead of chi square since outcome is Likert-scale item (i.e., ordinal variable)

visualize with either mosaic plots or boxplots

Table of kruskal wallis test results

Q2_item	Q2_item_name	statistic	p.value	parameter	method
Q2a	Private/Corporate	11.902255	0.0180931	4	Kruskal-Wallis rank sum test
Q2b	Non-profit/NGO	38.804872	0.0000001	4	Kruskal-Wallis rank sum test
Q2c	Gov./Public Policy	7.208656	0.1252641	4	Kruskal-Wallis rank sum test
Q2d	Education	20.059687	0.0004860	4	Kruskal-Wallis rank sum test
Q2e	Entrepreneurship/Start-up	3.169778	0.5298262	4	Kruskal-Wallis rank sum test
Q2f	Healthcare	1.945414	0.7457981	4	Kruskal-Wallis rank sum test
Q2g	Other	13.747659	0.0081455	4	Kruskal-Wallis rank sum test

Clusters and Question 3 items (Pursuing various degree options)

Chi square tests of whether different clusters (students with different temporal discounting patterns of when climate change will affect various groups) want to pursue various degree options

## # A tibble: 18,256 x 5
##    student_id major cluster_time_rank Q3_item Q3_resp
##         <int> <chr>             <int> <chr>     <dbl>
##  1          1 Ele                   5 Q3a           0
##  2          1 Ele                   5 Q3b           1
##  3          1 Ele                   5 Q3c           0
##  4          1 Ele                   5 Q3d           0
##  5          1 Ele                   5 Q3e           0
##  6          1 Ele                   5 Q3f           0
##  7          1 Ele                   5 Q3g           0
##  8          2 Ele                   4 Q3a           0
##  9          2 Ele                   4 Q3b           1
## 10          2 Ele                   4 Q3c           0
## # ... with 18,246 more rows

Interest of each topic

Interest within each cluster

Chi square tests of degree interests by cluster – not expecting any differences across clusters

Q3_item	Q3_item_name	statistic	p.value	parameter	method
Q3a	MA/MS (non-eng)	20.934492	0.0018850	6	Pearson’s Chi-squared test
Q3b	ME/MS (eng)	17.392641	0.0079436	6	Pearson’s Chi-squared test
Q3c	PhD (eng)	8.725628	0.1896066	6	Pearson’s Chi-squared test
Q3d	MBA	5.858231	0.4392572	6	Pearson’s Chi-squared test
Q3e	JD (law)	10.819632	0.0941135	6	Pearson’s Chi-squared test
Q3f	MD	5.223481	0.5154862	6	Pearson’s Chi-squared test
Q3g	Other	8.527301	0.2019589	6	Pearson’s Chi-squared test

Kruskal wallis tests of whether different clusters (students with different temporal discounting patterns of when climate change will affect various groups) consider various career satisfaction factors important

Career satisfaction items (Q4 items)

## # A tibble: 41,728 x 5
##    student_id major cluster_time_rank Q4_item Q4_resp
##         <int> <chr>             <int> <chr>     <dbl>
##  1          1 Ele                   5 Q4a           4
##  2          1 Ele                   5 Q4b           3
##  3          1 Ele                   5 Q4c           3
##  4          1 Ele                   5 Q4d           2
##  5          1 Ele                   5 Q4e           4
##  6          1 Ele                   5 Q4f           2
##  7          1 Ele                   5 Q4g           2
##  8          1 Ele                   5 Q4h           4
##  9          1 Ele                   5 Q4i           4
## 10          1 Ele                   5 Q4j           1
## # ... with 41,718 more rows

Table of chi square test results

Q4_item	Q4_item_name	statistic	p.value	parameter	method
Q4a	Make money	27.77257	0.2697493	24	Pearson’s Chi-squared test
Q4b	Fame	23.05203	0.5167281	24	Pearson’s Chi-squared test
Q4c	Help others	59.77527	0.0000687	24	Pearson’s Chi-squared test
Q4d	Supervise others	24.54317	0.4309141	24	Pearson’s Chi-squared test
Q4e	Job sec. and opp.	41.83517	0.0134640	24	Pearson’s Chi-squared test
Q4f	Work w/ people	49.69039	0.0015501	24	Pearson’s Chi-squared test
Q4g	Invent/design	33.19928	0.0999380	24	Pearson’s Chi-squared test
Q4h	Develop knowledge/skill	30.05447	0.1829529	24	Pearson’s Chi-squared test
Q4i	Personal/fam. time	30.50092	0.1686997	24	Pearson’s Chi-squared test
Q4j	Easy job	72.65467	0.0000009	24	Pearson’s Chi-squared test
Q4k	Exciting env.	29.78337	0.1920356	24	Pearson’s Chi-squared test
Q4l	Solve societal prob.	94.89931	0.0000000	24	Pearson’s Chi-squared test
Q4m	Use talent/abilities	37.59814	0.0381002	24	Pearson’s Chi-squared test
Q4n	Do hands-on work	27.20240	0.2951075	24	Pearson’s Chi-squared test
Q4o	Apply math/sci.	26.14879	0.3456586	24	Pearson’s Chi-squared test
Q4p	Volunteer w/ charity	78.86848	0.0000001	24	Pearson’s Chi-squared test

Using kruskal-wallis instead of chi square since outcome is Likert-scale item (i.e., ordinal variable)

Q4_item	Q4_item_name	statistic	p.value	parameter	method
Q4a	Make money	12.819453	0.0121925	4	Kruskal-Wallis rank sum test
Q4b	Fame	2.078992	0.7212328	4	Kruskal-Wallis rank sum test
Q4c	Help others	25.037874	0.0000494	4	Kruskal-Wallis rank sum test
Q4d	Supervise others	2.835811	0.5856670	4	Kruskal-Wallis rank sum test
Q4e	Job sec. and opp.	12.838757	0.0120911	4	Kruskal-Wallis rank sum test
Q4f	Work w/ people	3.993174	0.4069304	4	Kruskal-Wallis rank sum test
Q4g	Invent/design	2.182131	0.7023020	4	Kruskal-Wallis rank sum test
Q4h	Develop knowledge/skill	7.608165	0.1070332	4	Kruskal-Wallis rank sum test
Q4i	Personal/fam. time	3.142947	0.5341950	4	Kruskal-Wallis rank sum test
Q4j	Easy job	5.616663	0.2296635	4	Kruskal-Wallis rank sum test
Q4k	Exciting env.	13.322136	0.0098045	4	Kruskal-Wallis rank sum test
Q4l	Solve societal prob.	74.868555	0.0000000	4	Kruskal-Wallis rank sum test
Q4m	Use talent/abilities	20.300904	0.0004355	4	Kruskal-Wallis rank sum test
Q4n	Do hands-on work	5.069584	0.2802320	4	Kruskal-Wallis rank sum test
Q4o	Apply math/sci.	1.734258	0.7844857	4	Kruskal-Wallis rank sum test
Q4p	Volunteer w/ charity	27.906034	0.0000130	4	Kruskal-Wallis rank sum test

Q5 career topic interests for all ten topics in Q5 by cluster (instead of major)

Table of average number of topics that students in each cluster identified

## # A tibble: 7 x 3
##   cluster_time_rank     n avg_Q5_total
##               <int> <int>        <dbl>
## 1                 1  1080         3.68
## 2                 5   126         3.21
## 3                 2   147         3.18
## 4                 4   264         2.98
## 5                 3   241         2.97
## 6                 6   575         2.71
## 7                 7   175         2.06

Number of Q5 topics identified in each cluster

Proportions of each cluster interested in Q5 topics

Interests within each cluster

Q5_item	Q5_item_name	statistic	p.value	parameter	method
Q5a	Energy (supply/demand)	13.803697	0.0319075	6	Pearson’s Chi-squared test
Q5b	Disease	7.383852	0.2868020	6	Pearson’s Chi-squared test
Q5c	Poverty and wealth dist.	52.194355	0.0000000	6	Pearson’s Chi-squared test
Q5d	Climate change	203.090564	0.0000000	6	Pearson’s Chi-squared test
Q5e	Terrorism and war	9.742207	0.1359365	6	Pearson’s Chi-squared test
Q5f	Water supply	27.878110	0.0000991	6	Pearson’s Chi-squared test
Q5g	Food availability	30.949248	0.0000259	6	Pearson’s Chi-squared test
Q5h	Opp. for future gen	15.781208	0.0149778	6	Pearson’s Chi-squared test
Q5i	Opp. for women and/or min.	85.960121	0.0000000	6	Pearson’s Chi-squared test
Q5j	Environmental degradation	117.482450	0.0000000	6	Pearson’s Chi-squared test

End of career interest section

End cluster analysis